Re: Cache Digest just doesn't work?

On Mon, 5 Oct 1998, Teemu Peltonen wrote:

> Level 5 was a bit too quiet (couldn't find any references in cache.log), but
> here's what I got with level 6:

That's impossible. Most of these messages are printed with level 3. Perhaps
you were looking at cache log when no digest activity was going on. Recall
that caches exchange digests once per hour only.
 
> 1998/10/05 14:36:22| peerDigestValidate: digest themis.netti.fi
> 1998/10/05 14:36:22| current GMT time: Mon, 05 Oct 1998 11:36:22 GMT
> 1998/10/05 14:36:22| peerDigestValidate: themis.netti.fi was disabled

"Disabled" is not very good, but it could be OK depending on the history.

> ...
> 1998/10/05 14:36:22| peerDigestRequest: forwarding to fwdStart...
> 1998/10/05 14:36:22| peerDigestFetchReply: themis.netti.fi status: 200, expires: Mon, 05 Oct 1998 12:24:44 GMT
> 1998/10/05 14:36:22| peerDigestFetchReply: got new digest, requesting release of old digest
> 1998/10/05 14:36:22| got digest cblock from themis.netti.fi; ver: 3 (req: 3)
> ...
> 1998/10/05 14:36:22| received valid digest from themis.netti.fi

This look OK.

> I don't see any problems there, but still there are no CACHE_DIGEST_HITs on
> access.log. Also, there are no error messages of any kind that would refer
> to other caches.

I would try requesting an object that is for sure in the peer cache:

  - Use client program that comes with Squid 2.0
  - Double check that the object is *cached* by your peer
    (e.g., use only-if-cached cache-control directive or HEAD method).
  - Purge the object from local cache (-m PURGE) if needed.
  - Request the object through your cache.
  - Analyze returned headers.

X-Cache should be MISS on your cache and HIT on peer cache. If
X-Cache-Lookup is HIT on peer cache, but X-Cache is MISS on peer cache, then
your peer decided to purge the object (refresh rules, etc).

While doing that you can also enable full debugging on your cache (or at
least for neighbor.c and peer_digest.c modules) to get more info.

If nothing helps, send headers and relevant debugging info back to us.
 
> > The most common problem is access controls for internal objects. There may
> > be other reasons, of course.
>
> What controls internal objects? I couldn't find a correct acl at
> documentation.

There are no special "internal acls". However, other acls may interfere with
internal objects. Most of such interference is force-fixed in Squid code, but
there might be some exceptions. However, your cache did receive a valid
digest. Thus, acls are not a problem; at least not for fetching digests.

Good luck,

Alex.

P.S. Be sure each peer has a unique name. If several caches have the same
name (RR DNS balancing and such), digests (and other things) may not work.
Received on Mon Oct 05 1998 - 14:37:19 MDT